17

I have an xarray DataArray that looks like this below with shape (1,5,73,144,17) and I'm trying to drop or delete the "level" coordinates. So, ultimately, i need the variable to have shape = (1,5,73,144).

stdna
Out[717]: 
<xarray.DataArray 'stack-6e9b86fc65e3f0fda2008a339e235bc7' (variable: 1, week: 5, lat: 73, lon: 144, 
level: 17)>
dask.array<stack, shape=(1, 5, 73, 144, 17), dtype=float32, chunksize=(1, 1, 73, 144, 17), 
chunktype=numpy.ndarray>
Coordinates:
* lon       (lon) float32 0.0 2.5 5.0 7.5 10.0 ... 350.0 352.5 355.0 357.5
* lat       (lat) float32 90.0 87.5 85.0 82.5 80.0 ... -82.5 -85.0 -87.5 -90.0
* level     (level) float32 1000.0 925.0 850.0 700.0 ... 50.0 30.0 20.0 10.0
* week      (week) int64 5 6 7 8 9
* variable  (variable) <U3 'hgt'

I've taken a look to the xarray documentation and it's not helping. I've tried different combinations around this idea but i usually get the statement below and the coordinate has not been removed:

s = stdna.drop('level', dim=None)

Dimensions without coordinates: level

Thank you for your help!

Improve this question

4 Answers 4

Reset to default
16

We can use the drop_vars method to drop a coord:

In [10]: da
Out[10]:
<xarray.DataArray (dim_0: 2, dim_1: 3)>
array([[0.15928504, 0.47081089, 0.50490985],
       [0.6151981 , 0.41735643, 0.2576089 ]])
Coordinates:
    x        (dim_0, dim_1) float64 0.1593 0.4708 0.5049 0.6152 0.4174 0.2576
Dimensions without coordinates: dim_0, dim_1

In [11]: da.drop_vars('x')
Out[11]:
<xarray.DataArray (dim_0: 2, dim_1: 3)>
array([[0.15928504, 0.47081089, 0.50490985],
       [0.6151981 , 0.41735643, 0.2576089 ]])
Dimensions without coordinates: dim_0, dim_1

Alternatively reset_coords('level', drop=True) will also work.

As discussed in the comments — if we want to reduce the size of an array, then we need to do some slicing or reduction operation. Check out @MichaelDelgado's answer for more details.

Improve this answer
Sign up to request clarification or add additional context in comments.

5 Comments

Note that these don’t actually reshape the array as requested, they merely delete the coordinate labels for level, resulting in a "dimension without coordinates"
I agree it doesn't reshape the array. But how would that work? If we have a 2x2x2 array what happens to the data such that we can get a 2x2 array? Presumably that requires an aggregation or a slice...
Ah, I see your answer @MichaelDelgado — very much agree. I think my answer is still helpful to answer "how do I drop a coord", but yours is a more valid response to the detailed question...
Yep this is super valuable - just flagging for those who come along with the same question - I’ve seen this ask to “drop a dim” when people really mean select or reduce :)
Yes perfect, thank you @MichaelDelgado! I'll edit the answer slightly
4

Late to the party, but in case anyone else comes across this...

The reason you can't drop a dimension like this is because your data is actually indexed by level. In order to "remove" the the level dimension from your data, you need to decide how you would like to reduce the information along this dimension.

You could do this in all kinds of ways. If you want to select a single level from the array, then da.sel is what you're looking for, e.g.:

stdna.sel(level=1000)

On the other hand, maybe you're looking to aggregate the data across the level dimension? For example, you could take the average value across all levels:

stdna.mean(dim='level')

But without knowing how you want to go from a DataArray that is indexed by level to one that is not, there's no way for xarray to simply "drop" it from the data - the array will still be shaped (1,5,73,144,17).

See the docs on indexing and selecting data or computation: aggregation for more info on these topics.

Improve this answer

1 Comment

If what you want is select a specific value along a dimension and then forget about the coordinate label (e.g. to prevent dummy data being added when combining different arrays into a dataset where each array was selected for a different value of the relevant coordinate), stdna.sel(level=1000).drop_vars(name="level") does that (without nullifying the selected data).
3

Initially, I looked at the behaviour of drop and found that it does not delete the dimension. It can be used to delete data variables as such.

Then I tried this:

del stdna['level']

I would say the best way would be to try :

stdna.drop_dims('level')

There is one more thing I tried :

stdna = stdna.drop([i for i in stdna.coords if i not in stdna.dims])

to see if I can generalize this issue. But don't think this would work well. From docs : http://xarray.pydata.org/en/stable/generated/xarray.Dataset.drop_dims.html

Improve this answer

1 Comment

Note that drop_dims will drop the entire array from the dataset, because the level dim is still a core dimension of the array. And you’re right that del and drop won’t do anything except remove the coordinate labels.
-1

If you have levels nested within the events of your DataArray like so:

Example da

You need to use .droplevel() within the MultiIndex. For this example, to drop the Coordinate answer_correct:

da.indexes['event'].droplevel('answer_correct')
Improve this answer

Comments

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.